Nvidia DGX Spark Update Boosts Efficiency with 32% Power Cut — Hot-Plug Detection Enhances AI Workstation Performance

The DGX Spark: A Compact AI Engine with Surprising Power Insights

Nvidia's DGX Spark has established itself as a powerful and versatile local AI engine, thanks to its 128GB of RAM, a fast 20-core Arm CPU, and a capable Blackwell GPU. During our review, we found it equally effective for large language model (LLM) inference and generative image and video workflows.

However, one unexpected finding emerged during our testing. Despite its advanced 3nm-class fabrication technology, the Arm CPU complex, and Nvidia's extensive experience in mobile power management, the DGX Spark and its GB10 SoC consumed around 37W at idle. While this isn't catastrophic for an otherwise efficient system, it raised questions about why the power draw was so high—and whether there was a solution.

In its most recent system software update, Nvidia introduced hot-plug detection for the Spark's 200Gbps ConnectX 7 NIC. This networking interface had previously been drawing significant power at idle. According to Nvidia, the update can reduce system power consumption by up to 18W when the ConnectX 7 interface is inactive.

To verify these claims, we used a USB-C power meter to measure the idle power consumption of the Founders Edition Spark before and after applying the latest update.

Before the update, the DGX Spark Founders Edition idled at approximately 37W, as observed in our initial review. After the update, the system drew just 25W at idle with the connected display on—a reduction of 32.4%. When the display was disconnected, simulating a headless server setup, the idle power dropped further to 22W.

Not all GB10 systems may benefit equally from this update. We tested the new software version on our Dell Pro Max GB10 system, but the idle power remained unchanged at 35-37W. We are still experimenting with the Pro Max and will continue to see if similar power reductions can be achieved on that system.

Nvidia is also highlighting how the DGX Spark is being utilized in educational institutions worldwide as a local AI accelerator for various innovative projects.

According to the company's latest blog post, DGX Sparks are now powering AI models at the IceCube Neutrino Observatory at the South Pole, processing radiology reports at NYU, and aiding in the study of epilepsy genetics at Harvard, among other applications.

Benedikt Riedel, computing director at the Wisconsin IceCube Particle Astrophysics Center, noted that the Spark's low power requirements make it ideal for deployment in the harsh and remote conditions of the South Pole for local analysis of neutrino observation data.

At the South Pole, network bandwidth and available power are limited, making it crucial to run AI models locally at high performance within a power budget of around 140W. The lab is likely to appreciate the additional power savings from the latest update.

If you own a DGX Spark, be sure to check the DGX Dashboard and navigate to the Settings tab to install the latest software update for yourself.